David Strömberg 
WH 2008 Analysis 

International Economic Studies Stockholm University SE106 91 Stockholm SWEDEN Phone: +46 8 16 43 77 Last updated, October 1, 2008 
The Bradley effectI have estimated the Bradley effect based on 431 elections for House, Senator and Governor, 19982006, containing 26 black candidates. The effect is 23 percent of the twoparty vote share. See Strömberg (2008) for details and discussion. Below are some election statistics in case the Bradley effect for Obama is of similar size as that estimated in the 19982006 races discussed above.
The table below show election statistics if Obama will suffer a Bradley effect of the same 23 percent size as the average (Democratic and Republican) candidates for the 22 House, Senate and Governor races 19982006 for which I have data.
November 3, black Democratic candidates The estimated Bradley effect for black Democratic candidates lowers Obama's win chances as of November 3 from 87 to 68 percent. The effect is is 1.2 percent with a standard error of 1.6, so not statistically different from zero. Still, the uncertainty about the size of the Bradley effect makes Obama's lead uncertain. I find this by using the full distribution of the estimated coefficient. I take a 100 draws from this distribution and estimate the average statistic.
The number below were computed using the 23 percent Bradley effect.
The polling data and vote data for elections with one black candidate is available here.



